from sklearn.decomposition import PCA

# ---------------------------------------------
# 最佳主成分数量
pca = PCA(n_components=None)
pca.fit_transform(x_train)
var_values = pca.explained_variance_ratio_

def select_best_components(var_, goal_var):
    best_components, total_var = 0, 0.0  # 初始化
    for var in var_:
        total_var += var
        best_components += 1
        if total_var >= goal_var:  # 达到目标方差，终止整个循环
            break
    return best_components
n = select_best_components(var_=var_values, goal_var=0.95)
print('\n>>>最好的特征数量为 %d 个'% n)

# ---------------------------------------------
# 根据上面的最佳 n_components 数量，进行降维
pca = PCA(n_components=10)
x_train_pca = pca.fit_transform(x_train)
x_test_pca = pca.fit_transform(x_test)

# ---------